Collaborative Research: WCR: Incorporation of Model Bias and Uncertainty in Land Surface Hydrologic Flux Prediction Using a Data Assimilation Framework

合作研究：WCR：使用数据同化框架将模型偏差和不确定性纳入陆地表面水文通量预测

• 批准号: 0333133
• 负责人: Steven Margulis
• 金额: $ 21.28万
• 依托单位: University of California-Los Angeles
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-06-01 至 2007-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0333133&HistoricalAwards=false
• 关键词: Collaborative; Research; WCR; Incorporation; Model

0333133MargulisA major weakness of current data assimilation algorithms is that they typically do not account for model errors or assume that the model errors are unbiased. However biases are inevitably introduced in the models due to poorly specified forcing and parameters. Understanding the variance and memory of these errors is key to properly including their representation in data assimilation frameworks and will ultimately lead to improving our ability to predict variations in hydrologic fluxes.In this project we will focus on incorporating biases and other errors that can occur as a result of erroneous forcing and parameterizations in a data assimilation framework. The first tasks in our proposed research project will consist of numerical simulations to characterize the types of errors that are specific to land surface hydrologic modeling. The goal is to develop parsimonious probabilistic error models (that account for biases and other model errors) that are necessary inputs to data assimilation algorithms. Variational data assimilation experiments will then be performed to design an observing system that can ultimately be incorporated into a real-time data assimilation framework and is capable of including these parameterizations of model biases and errors. The final set of research tasks will be to incorporate the error models and observing system into a real-time (operational) data assimilation framework using the Ensemble Kalman Filter.The expected results of this study include the characterization of those errors that are caused by the inevitable misspecification of surface forcing and model parameters and incorporate them into land data assimilation systems.

当前数据同化算法的一个主要弱点是它们通常不考虑模式误差或假设模式误差是无偏的。然而，由于强迫和参数的指定不当，模型中不可避免地引入了偏差。了解这些误差的变化和记忆是将它们适当地包括在数据同化框架中的关键，并最终将提高我们预测水文通量变化的能力。在这个项目中，我们将重点考虑在数据同化框架中由于错误的强迫和参数化而可能发生的偏差和其他误差。我们提议的研究项目的第一个任务将包括数值模拟，以表征陆地表面水文模拟特有的误差类型。目标是开发简约的概率误差模型(考虑到偏差和其他模型误差)，这些模型是数据同化算法的必要输入。然后将进行变分数据同化实验，以设计一个观测系统，该系统最终可纳入实时数据同化框架，并能够包括这些模型偏差和误差的参数。最后一组研究任务将是将误差模式和观测系统纳入使用集合卡尔曼过滤器的实时(业务)数据同化框架中。这项研究的预期结果包括表征由于不可避免地错误指定地面强迫和模式参数而造成的误差，并将其纳入陆地数据同化系统。